Lance is a modern columnar data format that is optimized for ML workflows and datasets. Lance is perfect for:

1. Building search engines and feature stores.
2. Large-scale ML training requiring high performance IO and shuffles.
3. Storing, querying, and inspecting deeply nested data for robotics or large blobs like images, point clouds, and more.

The key features of Lance include:

• High-performance random access: 100x faster than Parquet without sacrificing scan performance.

• Vector search: find nearest neighbors in milliseconds and combine OLAP-queries with vector search.

• Zero-copy, automatic versioning: manage versions of your data without needing extra infrastructure.

• Ecosystem integrations: Apache Arrow, Pandas, Polars, DuckDB, Ray, Spark and more on the way.

[!TIP] Lance is in active development and we welcome contributions. Please see our contributing guide for more information.

Quick Start

Installation

shell pip install pylance

To install a preview release:

shell pip install --pre --extra-index-url https://pypi.fury.io/lancedb/ pylance

[!TIP] Preview releases are released more often than full releases and contain the latest features and bug fixes. They receive the same level of testing as full releases. We guarantee they will remain published and available for download for at least 6 months. When you want to pin to a specific version, prefer a stable release.

Converting to Lance

```python import lance

import pandas as pd import pyarrow as pa import pyarrow.dataset

df = pd.DataFrame({"a": [5], "b": [10]}) uri = "/tmp/test.parquet" tbl = pa.Table.frompandas(df) pa.dataset.writedataset(tbl, uri, format='parquet')

parquet = pa.dataset.dataset(uri, format='parquet') lance.write_dataset(parquet, "/tmp/test.lance") ```

Reading Lance data python dataset = lance.dataset("/tmp/test.lance") assert isinstance(dataset, pa.dataset.Dataset)

Pandas python df = dataset.to_table().to_pandas() df

DuckDB ```python import duckdb

If this segfaults, make sure you have duckdb v0.7+ installed

duckdb.query("SELECT * FROM dataset LIMIT 10").to_df() ```

Vector search

Download the sift1m subset

shell wget ftp://ftp.irisa.fr/local/texmex/corpus/sift.tar.gz tar -xzf sift.tar.gz

Convert it to Lance

```python import lance from lance.vector import vectotable import numpy as np import struct

nvecs = 1000000 ndims = 128 with open("sift/sift_base.fvecs", mode="rb") as fobj: buf = fobj.read() data = np.array(struct.unpack("<128000000f", buf[4 : 4 + 4 * nvecs * ndims])).reshape((nvecs, ndims)) dd = dict(zip(range(nvecs), data))

table = vectotable(dd) uri = "vecdata.lance" sift1m = lance.writedataset(table, uri, maxrowspergroup=8192, maxrowsperfile=1024*1024) ```

Build the index

python sift1m.create_index("vector", index_type="IVF_PQ", num_partitions=256, # IVF num_sub_vectors=16) # PQ

Search the dataset

```python

Get top 10 similar vectors

import duckdb

dataset = lance.dataset(uri)

Sample 100 query vectors. If this segfaults, make sure you have duckdb v0.7+ installed

sample = duckdb.query("SELECT vector FROM dataset USING SAMPLE 100").todf() queryvectors = np.array([np.array(x) for x in sample.vector])

Get nearest neighbors for all of them

rs = [dataset.totable(nearest={"column": "vector", "k": 10, "q": q}) for q in queryvectors] ```

Directory structure

| Directory | Description | |--------------------|--------------------------| | rust | Core Rust implementation | | python | Python bindings (PyO3) | | java | Java bindings (JNI) | | docs | Documentation source |

What makes Lance different

Here we will highlight a few aspects of Lance’s design. For more details, see the full Lance design document.

Vector index: Vector index for similarity search over embedding space. Support both CPUs (x86_64 and arm) and GPU (Nvidia (cuda) and Apple Silicon (mps)).

Encodings: To achieve both fast columnar scan and sub-linear point queries, Lance uses custom encodings and layouts.

Nested fields: Lance stores each subfield as a separate column to support efficient filters like “find images where detected objects include cats”.

Versioning: A Manifest can be used to record snapshots. Currently we support creating new versions automatically via appends, overwrites, and index creation.

Fast updates (ROADMAP): Updates will be supported via write-ahead logs.

Rich secondary indices: Support BTree, Bitmap, Full text search, Label list, NGrams, and more.

Benchmarks

Vector search

We used the SIFT dataset to benchmark our results with 1M vectors of 128D

1. For 100 randomly sampled query vectors, we get <1ms average response time (on a 2023 m2 MacBook Air)

1. ANNs are always a trade-off between recall and performance

Vs. parquet

We create a Lance dataset using the Oxford Pet dataset to do some preliminary performance testing of Lance as compared to Parquet and raw image/XMLs. For analytics queries, Lance is 50-100x better than reading the raw metadata. For batched random access, Lance is 100x better than both parquet and raw files.

Why are you building yet another data format?!

The machine learning development cycle involves the steps:

mermaid graph LR A[Collection] --> B[Exploration]; B --> C[Analytics]; C --> D[Feature Engineer]; D --> E[Training]; E --> F[Evaluation]; F --> C; E --> G[Deployment]; G --> H[Monitoring]; H --> A;

People use different data representations to varying stages for the performance or limited by the tooling available. Academia mainly uses XML / JSON for annotations and zipped images/sensors data for deep learning, which is difficult to integrate into data infrastructure and slow to train over cloud storage. While industry uses data lakes (Parquet-based techniques, i.e., Delta Lake, Iceberg) or data warehouses (AWS Redshift or Google BigQuery) to collect and analyze data, they have to convert the data into training-friendly formats, such as Rikai/Petastorm or TFRecord. Multiple single-purpose data transforms, as well as syncing copies between cloud storage to local training instances have become a common practice.

While each of the existing data formats excels at the workload it was originally designed for, we need a new data format tailored for multistage ML development cycles to reduce and data silos.

A comparison of different data formats in each stage of ML development cycle.

| | Lance | Parquet & ORC | JSON & XML | TFRecord | Database | Warehouse | |---------------------|-------|---------------|------------|----------|----------|-----------| | Analytics | Fast | Fast | Slow | Slow | Decent | Fast | | Feature Engineering | Fast | Fast | Decent | Slow | Decent | Good | | Training | Fast | Decent | Slow | Fast | N/A | N/A | | Exploration | Fast | Slow | Fast | Slow | Fast | Decent | | Infra Support | Rich | Rich | Decent | Limited | Rich | Rich |

Community Highlights

Lance is currently used in production by: * LanceDB, a serverless, low-latency vector database for ML applications * LanceDB Enterprise, hyperscale LanceDB with enterprise SLA. * Leading multimodal Gen AI companies for training over petabyte-scale multimodal data. * Self-driving car company for large-scale storage, retrieval and processing of multi-modal data. * E-commerce company for billion-scale+ vector personalized search. * and more.

Presentations, Blogs and Talks

• Designing a Table Format for ML Workloads, Feb 2025.
• Transforming Multimodal Data Management with LanceDB, Ray Summit, Oct 2024.
• Lance v2: A columnar container format for modern data, Apr 2024.
• Lance Deep Dive. July 2023.
• Lance: A New Columnar Data Format, Scipy 2022, Austin, TX. July, 2022.